TECHNIQUES IN INTRINSIC ANALYSIS.
Final rept., 1 Nov 65-31 Oct 66,
SYSTEMS RESEARCH LABS INC DAYTON OH
Pagination or Media Count:
Intrinsic analysis is a data reduction technique which allows a set of data vectors to be approximated to a given mean square error by a minimum number of coefficients. This report develops the analysis, showing the relation to other developments and proves several results about the relation between the row space and column space of matrices. The problem of error propagation and control is considered from several points of view and the effect of ambiguous data on the algebraic eigenvalue problem discussed. Finally, a survey is made of computational algorithms for the algebraic eigenproblem for large ambiguous second momet matrices, and two new algorithms proposed one for very large matrices and the other for complex hermitian matrices.
- Statistics and Probability